🧠 Deep Research AI - Advanced Multi-Agent Research System

An advanced AI-powered research system combining multi-agent architecture with cutting-edge reasoning techniques

Features • Architecture • Installation • Usage • Deployment • Screenshots

---

📖 Table of Contents

• Overview
• Features
• System Architecture
• Technology Stack
• Installation
• Configuration
• Usage
• API Documentation
• Screenshots
• Deployment
• Project Structure
• Contributing
• License

---

🌟 Overview

Deep Research AI is a sophisticated research automation system that leverages advanced AI reasoning techniques to conduct comprehensive, multi-dimensional research on any topic. Built with a multi-agent architecture, it performs parallel web searches, cross-validates sources, and generates detailed, citation-rich reports.

Why Deep Research AI?

• 🤖 Multi-Agent Architecture: Deploys 3+ specialized agents working in parallel
• 🧠 Advanced Reasoning: Implements Chain of Thought, Self-Consistency, and Tree of Thoughts
• ⭐ Source Verification: Cross-validates information across multiple sources
• 📊 Comprehensive Reports: Generates executive summaries with full citations
• 🎨 Beautiful UI: Modern, responsive interface with real-time progress tracking
• 🚀 Production Ready: Fully functional and deployable system

---

✨ Features

Core Capabilities

• 🔍 Intelligent Query Decomposition

  • Chain of Thought reasoning for query analysis
  • Automatic breakdown into research dimensions
  • Strategic search query generation

• 🤖 Multi-Agent Parallel Execution

  • 3 specialized agents per research task
  • Asynchronous parallel web searches
  • Real-time progress tracking

• ✅ Self-Consistency Validation

  • Cross-validation of sources
  • Confidence scoring based on repeated mentions
  • Deduplication and aggregation

• 📝 Advanced Report Generation

  • AI-powered executive summaries
  • Detailed sectional analysis
  • Proper citation formatting
  • Export to Markdown and JSON

• ⭐ Source Verification

  • High-confidence source identification
  • Cross-reference tracking
  • Verification badges

Technical Features

• FastAPI backend with async support
• React-based responsive frontend
• RESTful API architecture
• Real-time progress updates
• Comprehensive error handling
• Configurable research depth (3-10 tasks)

---

🏗️ System Architecture

Workflow Diagram

%%{init: {
  "theme": "base",
  "themeVariables": {
    "primaryColor": "#ffffff",
    "primaryTextColor": "#1f2937",
    "lineColor": "#9ca3af",
    "edgeLabelBackground":"#ffffff",
    "fontSize": "14px",
    "fontFamily": "Inter, sans-serif",
    "tertiaryColor": "#f3f4f6"
  }
}}%%

graph TB
    Start(["🔍 User Research Query"]) --> Step1

    %% STEP 1
    subgraph Step1["Step 1 ▸ Query Analysis & Planning"]
        A["🧩 User Query Input"] --> B["🧠 Chain of Thought<br/>Decomposition"]
        B --> C["🔎 Query Analysis"]
        C --> D["🗂️ Generate Research Dimensions"]
        D --> E["🌳 Create Sub-Queries (Tree of Thoughts)"]
    end

    Step1 --> Step2

    %% STEP 2
    subgraph Step2["Step 2 ▸ Multi-Agent Execution"]
        E --> F1["🤖 Agent 1<br/>Market Analysis"]
        E --> F2["🤖 Agent 2<br/>Competition Research"]
        E --> F3["🤖 Agent 3<br/>Technical Analysis"]
        
        F1 --> G1["🌐 Web Search 1"]
        F1 --> G2["🌐 Web Search 2"]
        F1 --> G3["🌐 Web Search 3"]
        
        F2 --> H1["🌐 Web Search 1"]
        F2 --> H2["🌐 Web Search 2"]
        F2 --> H3["🌐 Web Search 3"]
        
        F3 --> I1["🌐 Web Search 1"]
        F3 --> I2["🌐 Web Search 2"]
        F3 --> I3["🌐 Web Search 3"]
        
        G1 --> J1["📄 Aggregated Results (Agent 1)"]
        H1 --> J2["📄 Aggregated Results (Agent 2)"]
        I1 --> J3["📄 Aggregated Results (Agent 3)"]
    end

    Step2 --> Step3

    %% STEP 3
    subgraph Step3["Step 3 ▸ Synthesis & Validation"]
        J1 --> K["📊 Aggregate All Results"]
        J2 --> K
        J3 --> K
        K --> L["🧩 Self-Consistency Validation"]
        L --> M["🧹 Deduplicate Sources"]
        M --> N["📈 Confidence Scoring"]
        N --> O["🔁 Cross-Validation"]
        O --> P["💡 Extract Key Insights"]
        P --> Q["🧠 LLM Synthesis (Sequential Revision)"]
        Q --> R["🗒️ Generate Executive Summary"]
        R --> S["📚 Create Detailed Sections"]
    end

    Step3 --> Step4

    %% STEP 4
    subgraph Step4["Step 4 ▸ Final Output"]
        S --> T["🧾 Format Citations"]
        T --> U["✅ Verification Report"]
        U --> V["📘 Complete Research Report"]
        V --> W1["⬇️ Download Markdown"]
        V --> W2["⬇️ Download JSON"]
        V --> W3["🌐 View in Browser"]
    end

    W3 --> End(["🎯 User Receives Results"])

    %% STYLE DEFINITIONS
    style Start fill:#e0f2fe,stroke:#0284c7,stroke-width:2px,color:#0c4a6e
    style End fill:#dcfce7,stroke:#16a34a,stroke-width:2px,color:#064e3b

    style Step1 fill:#e0f2fe,stroke:#0284c7,stroke-width:2px,color:#0c4a6e
    style Step2 fill:#ede9fe,stroke:#7c3aed,stroke-width:2px,color:#312e81
    style Step3 fill:#f0fdf4,stroke:#22c55e,stroke-width:2px,color:#064e3b
    style Step4 fill:#fff7ed,stroke:#f97316,stroke-width:2px,color:#7c2d12

Loading

Advanced Reasoning Techniques

The system implements multiple state-of-the-art AI reasoning approaches:

1. Chain of Thought (CoT) - Step-by-step problem decomposition
2. Tree of Thoughts (ToT) - Exploration of multiple reasoning paths
3. Self-Consistency - Multiple sampling with majority voting
4. Sequential Revision - Iterative refinement of outputs
5. Search & Verify - Solution space exploration with verification

---

🛠️ Technology Stack

Backend

• Framework: FastAPI 0.104+
• Language: Python 3.8+
• Async: asyncio, aiohttp
• API Integration:
  • OpenRouter API (LLM)
  • SerpAPI (Web Search)

Frontend

• Library: React 18
• Styling: Tailwind CSS + Custom CSS
• Build: Babel (in-browser compilation)
• HTTP: Fetch API

APIs & Services

• OpenRouter: GPT-4o-mini, Claude 3.5 Sonnet, Llama 3.1
• SerpAPI: Google search results

---

📦 Installation

Prerequisites

• Python 3.8 or higher
• pip (Python package manager)
• OpenRouter API key (Get here)
• SerpAPI key (Get here)

Backend Setup

# Clone the repository
git clone https://github.com/AdiityaRaj/deep-research-ai.git
cd deep-research-ai

# Create virtual environment
python -m venv venv

# Activate virtual environment
# On Windows:
venv\Scripts\activate
# On macOS/Linux:
source venv/bin/activate

# Install dependencies
cd backend
pip install -r requirements.txt

# Create .env file
cp .env.example .env
# Edit .env and add your API keys

Frontend Setup

# Navigate to frontend directory
cd ../frontend

# No build required! Just open index.html
# Or use a local server:
python -m http.server 8080
           or
python -m http.server 8080 --bind 127.0.0.1 

---

⚙️ Configuration

Environment Variables

Create a .env file in the backend/ directory:

# OpenRouter Configuration
OPENROUTER_API_KEY=sk-or-v1-your-key-here
MODEL_NAME=openai/gpt-4o-mini

# SerpAPI Configuration
SERPAPI_KEY=your-serpapi-key-here

# Optional: Advanced Settings
MAX_PARALLEL_AGENTS=3
SEARCHES_PER_AGENT=3
CONFIDENCE_THRESHOLD=2

Frontend Configuration

Edit frontend/js/utils.js to configure API endpoint:

const API_CONFIG = {
    baseUrl: 'http://localhost:8000'  // Change for production
};

---

🚀 Usage

Starting the Application

1. Start Backend:

cd backend
uvicorn main:app --reload

Backend runs at: http://localhost:8000

2. Start Frontend:

cd frontend
python -m http.server 8080

Frontend runs at: http://localhost:8080

Running Research

Via Web Interface:

1. Open http://localhost:8080 in browser
2. Enter research query or select an example
3. Adjust research depth (3-10 tasks)
4. Click "Start Research"
5. Wait 30-90 seconds for results
6. Download report (Markdown or JSON)

Via API:

curl -X POST http://localhost:8000/run \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Analyze AI agent trends in 2024",
    "max_tasks": 5
  }'

Example Queries

Try these sample queries:

• "Analyze the competitive landscape of AI agents in 2024"
• "Latest developments in quantum computing"
• "Electric vehicle market trends 2024"
• "Impact of generative AI on software development"
• "Future of renewable energy technologies"

---

📚 API Documentation

Endpoints

GET /

Health check endpoint

{
  "status": "operational",
  "version": "4.0",
  "features": [...]
}

GET /health

Detailed system status

{
  "status": "healthy",
  "openrouter_configured": true,
  "serpapi_configured": true,
  "model": "openai/gpt-4o-mini"
}

POST /run

Execute deep research

Request:

{
  "query": "Your research question",
  "max_tasks": 5
}

Response:

{
  "query": "...",
  "planning": {
    "dimensions": [...],
    "total_searches": 9
  },
  "executive_summary": "...",
  "sections": [...],
  "all_sources": [...],
  "verification": [...],
  "metadata": {
    "total_sources": 42,
    "high_confidence_sources": 5,
    "techniques_used": [...]
  }
}

Interactive API Docs

Once backend is running, visit:

• Swagger UI: http://localhost:8000/docs
• ReDoc: http://localhost:8000/redoc

---

📸 Screenshots

Main Interface

Beautiful landing page with example queries and animated gradient background

Research in Progress

Real-time progress tracking with stage updates and animated progress bar

Results Dashboard

Comprehensive results with metrics, agent strategy, and sections

Executive Summary

AI-generated comprehensive summary of research findings

Detailed Sections

Expandable sections with key findings and source citations

Source Verification

Sources with confidence scores and verification badges

Research Strategy

Multi-agent deployment showing parallel research dimensions

Full web Page

Multi-agent deployment showing full web page of deep research ai

🌐 Deployment

Backend Deployment

Railway (Recommended):

cd backend
railway login
railway init
railway up

Render:

1. Connect GitHub repository
2. Select backend/ as root directory
3. Build command: pip install -r requirements.txt
4. Start command: uvicorn main:app --host 0.0.0.0

Fly.io:

cd backend
fly launch
fly deploy

Frontend Deployment

Vercel (Recommended):

cd frontend
vercel

Netlify:

1. Drag and drop frontend/ folder to Netlify
2. Or connect GitHub repository

GitHub Pages:

cd frontend
git subtree push --prefix frontend origin gh-pages

Environment Variables for Production

Update .env in backend and js/utils.js in frontend with production URLs.

---

📁 Project Structure

Deep-Research-AI/
├── backend/
│   ├── main.py                    # FastAPI application
│   ├── research_agent.py          # Main orchestrator
│   ├── query_decomposer.py        # Chain of Thought
│   ├── multi_agent_executor.py    # Multi-agent execution
│   ├── synthesizer.py             # Self-Consistency & synthesis
│   ├── citation_manager.py        # Citation formatting
│   ├── config.py                  # Configuration
│   ├── requirements.txt           # Python dependencies
│   └── .env                       # Environment variables
│
├── frontend/
│   ├── index.html                 # Main HTML
│   ├── css/
│   │   └── styles.css            # Custom styles
│   ├── js/
│   │   ├── utils.js              # Utilities
│   │   ├── components.js         # React components
│   │   └── app.js                # Main app
│   └── assets/                    # Images/icons
│
├── docs/
│   ├── shscreenots/              # UI screenshots
│   └── architecture.md           # Architecture docs
│
├── .gitignore
├── LICENSE
└── README.md                      # This file

---

🧪 Testing

Backend Tests

cd backend
python test_system.py

Expected output:

✅ Health check passed
✅ Configuration check passed
✅ Research completed in 85.44s
✅ ALL TESTS PASSED!

Manual Testing Checklist

• Backend starts without errors
• Frontend loads with animations
• Can submit research query
• Progress bar animates
• Results display correctly
• Sources show confidence scores
• Download functionality works
• Mobile responsive

---

🔧 Troubleshooting

Common Issues

Backend won't start:

• Check Python version (3.8+)
• Verify all dependencies installed
• Check .env file exists with valid API keys

Frontend shows blank page:

• Check browser console for errors
• Verify all 5 files created correctly
• Check backend is running

API connection failed:

• Verify backend URL in js/utils.js
• Check CORS is enabled in backend
• Ensure backend is running on correct port

Search results empty:

• Verify SerpAPI key is valid
• Check API quota/limits
• Try different query

---

🤝 Contributing

Contributions are welcome! Please follow these steps:

1. Fork the repository
2. Create feature branch (git checkout -b feature/amazing-feature)
3. Commit changes (git commit -m 'Add amazing feature')
4. Push to branch (git push origin feature/amazing-feature)
5. Open Pull Request

Development Guidelines

• Follow PEP 8 for Python code
• Use meaningful variable names
• Add comments for complex logic
• Test before submitting PR
• Update documentation as needed

---

📊 Performance

• Query Processing: 30-90 seconds average
• Parallel Searches: 9 searches per query (3 agents × 3 searches)
• Source Analysis: 40+ sources per research
• API Calls: ~12 total (9 SerpAPI + 3 OpenRouter)

Optimization Tips

• Reduce max_tasks for faster results
• Use caching for repeated queries
• Adjust CONFIDENCE_THRESHOLD for stricter validation
• Configure MAX_PARALLEL_AGENTS based on API limits

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

---

🙏 Acknowledgments

• OpenRouter - LLM API access
• SerpAPI - Web search functionality
• FastAPI - Backend framework
• React - Frontend library
• Tailwind CSS - Styling framework

---

📞 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: rajaditya2424@gmail.com

---

🗺️ Roadmap

Version 1.1 (Upcoming)

• WebSocket support for real-time updates
• User authentication and query history
• Custom agent configuration
• PDF export functionality
• Batch research processing

Version 2.0 (Future)

• Local LLM support (Ollama)
• Advanced visualization charts
• Collaborative research features
• Mobile app (React Native)
• Plugin system for extensibility

---

⭐ Star History

If you find this project useful, please consider giving it a star!

---

Built with ❤️ by Aditya Raj

⬆ Back to Top